About

I am a Postdoctoral Fellow at ASTAPLE, where I am honored to work with Prof. Haibo HU, in Department of Electrical and Electronic Engineering of The Hong Kong Polytechnic University.

Previously, I completed my Ph.D. in March 2025 under the supervision of Prof. Jie YANG at Institute of Image Processing and Pattern Recognition in Department of Automation of Shanghai Jiao Tong University. Before that, I received the BA.Eng. degree from Tongji University in 2018. I then received the MA.Eng. degree from Shanghai Jiao Tong University in 2021.

Research interests

My research interests focus on trustworthy deep learning. Currently, my works are about the robustness and privacy aspects of Deep Neural Networks (DNNs), within the fields of out-of-distribution detection and machine unlearning.

• Out-of-distribution detection: DNNs are typically trained on a fixed In-Distribution (InD) and tend to produce overconfident predictions on inputs that deviate from this distribution. My research aims to design detection methods that reliably distinguish such out-of-distribution samples from InD data, thereby enhancing model robustness in open-world deployment.
• Machine unlearning: Trained models often retain traces of their training data, raising privacy concerns. Machine unlearning tackles this by efficiently removing the influence of specific data points upon request, eliminating the need for costly full retraining. My work explores unlearning algorithms that balance data removal effectiveness with model fidelity, strengthening the privacy guarantees of DNNs.

A central theme of my research is the exploration of low-dimensional structures in data representations and model parameters. This perspective provides a principled lens to analyze, detect, and manipulate the critical properties underlying trustworthy deep learning, aiming for more interpretable, efficient, and fundamentally grounded solutions.

In addition, my earlier research also touched on kernel methods and adversarial robustness.

Awards

• Gold Reviewer for ICML 2026
• Outstanding Reviewer for ECCV 2024.

Last update: 2026.06.20.